On Thursday, August 7, 2003, at 06:02 AM, Alexandru Petrescu wrote:
S Woodside wrote:Correct me if I'm wrong, the IETF doesn't have any activities looking at the fixed mesh case.
If you were to look into fixed meshes, how would you approach the problem definition and what tools would you use? My approach to it would be to use address autoconfiguration coupled with routing protocol exchanges. The deployment space would be rather limited, say no larger than 50 vertices for the mesh.
I would not want to lock into address autoconfiguration, assigned IP address should be possible as well. 50 nodes is a little low, it would be nice to have hundreds, maybe thousands, at the "metropolitan" level. Hundreds to start with would suffice, if later improvements would show promise to scale that up. AODV breaks down over a few hundred IIRC.
It seems to me that gateways to the internet are important. Ideally , there would be an autoconfiguration option, a multihoming option (e.g. the mesh is multihomed) and a relatively simple managed option. I would like to see a situation where an ad-hoc mesh can also have ad-hoc gateways to the internet. After all the fixed mesh is not just for contacting other mesh nodes, but to get to the internet too ;-)
simon